RESEARCH

ELECTROCHEMISTRY the oxidation of 5-hydroxymethylfurfural (15–17), alcohol (18–20), and glycerol (21–23)arecon- Chloride-mediated selective electrosynthesis of ducted at low current densities (<100 mA/cm2) to maximize Faradaic efficiencies toward the and propylene oxides at high target product (Fig. 1C). However, the produc- tion of industrially relevant quantities of the current density product at such low current densities would require unreasonably high electrolyzer surface Wan Ru Leow1*, Yanwei Lum1,2*, Adnan Ozden3, Yuhang Wang1, Dae-Hyun Nam1, Bin Chen1, areas, leading to high capital costs per unit of Joshua Wicks1, Tao-Tao Zhuang1, Fengwang Li1, David Sinton3, Edward H. Sargent1† productivity (Fig. 1D). Second, if the reactant (ethylene) has limited solubility in the aqueous Chemicals consumes large amounts of energy and is responsible for a substantial electrolyte, the system quickly becomes mass portion of global carbon emissions. Electrochemical systems that produce the desired compounds by transport–limited, resulting in poor Faradaic using renewable electricity offer a route to lower carbon emissions in the chemicals sector. Ethylene efficiency at high current densities. oxide is among the world’s most abundantly produced commodity chemicals because of its importance We took the view that a selective production in the , notably for manufacturing polyesters and polyethylene terephthalates. We strategy could avoid directly oxidizing the or- applied an extended heterogeneous:homogeneous interface, using chloride as a redox mediator ganic reactant molecules on the electrode sur- at the anode, to facilitate the selective partial oxidation of ethylene to . We achieved face so as to prevent over-oxidation at high current densities of 1 ampere per square centimeter, Faradaic efficiencies of ~70%, and product current densities. We reasoned that a redox specificities of ~97%. When run at 300 milliamperes per square centimeter for 100 hours, the system mediator that facilitates the indirect exchange Downloaded from maintained a 71(±1)% Faradaic efficiency throughout. of electrons between the electrode and the substrate molecules would allow this. Further- more, in such a scheme, the space in which n the United States, chemical manufac- of the mechanism relied on in direct ethylene the reaction takes place is not limited to the ture accounts for 28% of total industrial oxidation: One ethylene reacts with one planar electrode:electrolyte interface but energy demand (1). At present, this de- (O) atom from Ag-adsorbed dioxygen to form extends into the bulk electrolyte, constituting I mand is largely met by the consumption ethylene oxide, leaving behind the remaining an extended heterogeneous:homogeneous in- http://science.sciencemag.org/ of fossil fuels, resulting in substantial car- Ag-adsorbed O atom, and this oxidizes ethylene terface (Fig. 2A) that overcomes mass transport bon dioxide (CO2) emissions (2, 3); a recent all the way to CO2. It takes six O atoms for the limitations. Using this strategy, we demon- report showed that the alone complete combustion of one ethylene mole- strated ethylene oxide production at high cur- 2 releases 1.8 billion metric tons of CO2 per year cule; this means that, in the best case, for rent density (up to 1 A/cm ), Faradaic efficiency and that replacing fossil fuels–based produc- every six molecules of ethylene converted to (~70%), and product specificity (~97%). tion methods with ones powered with re- ethylene oxide, one molecule of ethylene will Initially, we attempted to oxidize ethylene newable energy offers a route to reduce net be completely combusted to CO2, thus limiting directly to ethylene oxide using a nanostruc- greenhouse gas emissions associated with the product specificity to a theoretical upper tured palladium (Pd) anode (fig. S2A). This plastics manufacture (4). limit of 85.7% (9). Here, specificity refers to the approach was based on a recent study in which

One attractive strategy involves the devel- percentage of reacted substrate (ethylene) olefins such as propylene were oxidized at low on June 11, 2020 opment of electrochemical systems that produce that goes toward the desired product (ethylene current densities (24) but did not translate to the necessary raw materials by using renewable oxide). the high current densities; at 300 mA/cm2,a electricity (5–8). Ethylene oxide is used in the Additionally, fossil-powered systems are typ- negligible Faradaic efficiency was obtained manufacture of plastics, detergents, thickeners, ically used to maintain a stable temperature toward ethylene oxide (fig. S2B). Operating at and solvents (9)andisamongtheworld’stop profile in order to suppress thermal runaway this high current density resulted in dissolution 15 most abundantly produced chemicals at in this highly exothermic reaction. of the Pd anode, as can be observed from the ~20 million metric tons per year (10, 11). At We pursued an electrochemical approach to rapidly increasing potential with time (fig. S2C). present, it is manufactured through the silver the production of ethylene oxide, to address Additionally, the use of organic mediators such (Ag)–catalyzed direct oxidation of ethylene at both the first problem (limited product spec- as TEMPO [(2,2,6,6-tetramethylpiperidin-1- high temperature and pressure (200° to 300°C ificity) and the second (the need to reduce the yl)oxyl] (15, 25) and NHPI (N-hydroxyphtha- and 1 to 3 MPa). This process generates 0.9 tons use of fossil energy sources in powering systems). limide) (26, 27)—amethodtoobtainhigh of CO2 per ton of ethylene oxide produced, with We sought a highly selective electrochemical selectivities for partial oxidation products at more than half attributed to the complete com- route, pursuing the electrooxidation of ethyl- the anode—failed in the generation of ethylene bustion of ethylene and the balance arising ene to ethylene oxide with high product speci- oxide and yielded instead only small amounts from temperature-control units (Fig. 1A), which ficity and Faradaic efficiency under ambient of acetate (Fig. 2B). today are fossil fuel–powered (12). conditions as a means to contribute to lowering Historically, ethylene oxide was produced The contribution from the complete oxida- CO2 emissions in the production of this chem- from ethylene through the addition of aque- tion of ethylene all the way to CO2 is a result ical (Fig. 1B) (13, 14). ous to form ethylene chlorohydrin, The electrosynthesis of ethylene oxide in- followed by dehydrochlorination with calcium 1Department of Electrical and Computer , volves the partial oxidation of ethylene, an to generate ethylene oxide (9). University of Toronto, 35 St. George Street, Toronto, ON anodic reaction. Reactions of this nature at However, the high cost associated with con- 2 M5S 1A4, Canada. Institute of Materials Research and high current density and Faradaic efficiency suming stoichiometric amounts of chlorine Engineering, Agency for Science, Technology and Research (A*STAR), 2 Fusionopolis Way, Innovis, Singapore 138634, are hampered by two challenges. First, the and hydroxide, as well as disposal of the water Singapore. 3Department of Mechanical and Industrial large positive potentials necessary can lead medium, has lowered interest in this process. Engineering, University of Toronto, 5 King’s College Road, to uncontrolled over-oxidation, generating We postulated that chloride (Cl–)couldbea Toronto, ON M5S 3G8, Canada. *These authors contributed equally to this work. undesired by-products such as CO2.Currently, redox mediator at the anode with extended †Corresponding author. Email: [email protected] reported anodic upgrading reactions such as heterogeneous:homogeneous interfaces. The

Leow et al., Science 368, 1228–1233 (2020) 12 June 2020 1of6 RESEARCH | REPORT Downloaded from http://science.sciencemag.org/

– – Fig. 1. Electrosynthesis of ethylene oxide by using renewable energy. (blue squares) (15 24, 34 36). Data for the system demonstrated in this work on June 11, 2020 (A and B) Schematics illustrating (A) the industrial thermochemical system and are shown for comparison (red square). (D) Breakdown of costs at current (B) the proposed electrochemical system. (C) Reported current densities and densities of 50 and 300 mA/cm2, as calculated from a TEA. TEA calculation Faradaic efficiencies for other anodic partial oxidation reactions in the literature details are provided in the supplementary text and fig. S1.

– – – Cl could thereby buffer ethylene from uncon- Experiments were first carried out at 300 Cathode: 2H2O+2e → H2 + 2OH (4) trolled oxidation and facilitate ethylene oxide mA/cm2. On the basis of prior studies (28), – production. in this process Cl is oxidized to Cl2 at the Pt – Mixing step: HOCH2CH2Cl + OH → This idea was tested in a flow-cell setup with anode (Eq. 1; e is the charge on the electron), – C H O+Cl +HO (5) 1.0 M potassium chloride (KCl) electrolyte, which disproportionates in the aqueous 2 4 2 in which ethylene was continuously sparged environment to form hypochlorous and hydro- – – into the anolyte, with platinum (Pt) foil as the chloric acid (HOCl and HCl, respectively) HCl + OH → Cl +H2O (6) working electrode (anode), nickel (Ni) foam as (Eq. 2) (29, 30); HOCl then reacts with ethyl- the counter electrode (cathode), and Ag/AgCl ene dissolved in the electrolyte to form eth- Overall: C2H4 +H2O → C2H4O+H2 (7) (3.0 M KCl) as the reference electrode (fig. S3). ylene chlorohydrin (HOCH2CH2Cl) (Eq. 3). Thegeometricsurfaceareaoftheanodeused Because HCl is not consumed, the pH of the was 1 cm2, and the catholyte and anolyte vol- anolyte becomes acidic over the course of the The final step involves addition of hydroxide umes were both 25 ml. An anion exchange mem- electrolysis (pH 1.1) (OH–),whichthenreactswithethylenechloro- brane (AEM) separated the anolyte and catholyte hydrin to yield the desired ethylene oxide and – 28 chambers. To determine the Faradaic efficiency, Anode: 2Cl– → Cl +2e– (1) regenerate Cl ( ). The hydrogen evolution product quantification was carried out by means 2 reaction (fig. S4) at the cathode during elec- of high-performance liquid chromatography trolysis generates the necessary OH– (Eq. 4), (HPLC) (supplementary materials, materials Cl2 þ H2O ⇋ HOCl þ HCl ð2Þ whereas the AEM prevents complete mixing and methods). Unless otherwise stated, all of the catholyte and the anolyte. Consequently, electrolysis experiments were run for a dura- at the end of electrolysis, the pH of the catho- C2H4 + HOCl → HOCH2CH2Cl (3) tion of 1 hour. lyte becomes alkaline, with a pH value of 13.8.

Leow et al., Science 368, 1228–1233 (2020) 12 June 2020 2of6 RESEARCH | REPORT Downloaded from http://science.sciencemag.org/

Fig. 2. Selective ethylene oxide production from ethylene enabled by an oxide and ethylene chlorohydrin. (D)Faradaicefficienciesofethyleneoxide extended heterogeneous:homogenous interface. (A) Schematic illustrating and ethylene chlorohydrin at different current densities. (E) Faradaic effi- ethylene oxidation at planar versus extended interfaces. (B) Schematic of ciencies of and propylene chlorohydrin at different current the ethylene-to–ethylene oxide electrochemical system. A detailed schematic densities. The error bars correspond to the standard deviation of three of the electrolyzer is available in fig. S3. (C) 13C NMR spectra of ethylene independent measurements.

on June 11, 2020

This means that by mixing the catholyte and Using this method, we achieved a Faradaic reactor process could be adopted (as opposed anolyte output streams (performed after elec- efficiency of 70(±1)% toward ethylene oxide to the current batch reactor mode), in which trolysis), ethylene oxide can be generated from (Fig. 2D) with 1.0 M KCl at 300 mA/cm2.This portions of the electrolyte are periodically the reaction between ethylene chlorohydrin corresponds to 3.9 mmol of ethylene oxide siphoned off for ethylene oxide extraction. and OH– (Eq. 5 and Fig. 2B). At the same time, produced after 1 hour of electrolysis. Similar In this case, the electrolyte would be contin- the HCl generated (Eq. 2) is also neutralized Faradaic efficiencies of 71(±1)% and 70(±1)% uously replenished by fresh or regenerated by OH– (Eq. 6). are maintained even at current densities of solution, preventing excessive buildup of these The formation of ethylene chlorohydrin in the 500 and 800 mA/cm2, respectively (Fig. 2D). corrosive species. anolyte and subsequent generation of ethyl- A possible explanation for the missing charge The method could also be used for the ep- ene oxide in the mixing step were confirmed could be O2 evolution or complete oxidation oxidations of other olefins; for example, when 1 through H nuclear magnetic resonance (NMR) of ethylene to form CO2; however, when we we replaced ethylene with propylene, Faradaic (fig. S5). We performed the same experiments performed gas chromatography on the output efficiencies were 69 to 71% toward propylene 13 but using carbon-13–labeled ethylene ( C2H4): gas stream, we did not detect O2 nor CO2.The oxide—a commodity chemical with a market of 13CNMRand1H NMR results confirm that the product specificity has a value of 97%; ethylene 10 million tons per year in the plastics industry 2 products we observed were indeed due to the conversion to other products (such as CO2)was (32)—at current densities of 300 to 800 mA/cm partial oxidation of ethylene (Fig. 2C and fig. not observed, and the remaining specificity (Fig. 2E). S5). In principle, a cation exchange membrane was due to incomplete conversion of ethylene We performed a technoeconomic analysis would be better in preventing crossover of OH– chlorohydrin to ethylene oxide (Eq. 5). (TEA) to identify conditions that could enable (fig. S6A). However, this would lead to a conti- We instead hypothesized that the missing the profitable synthesis of a renewable energy– nuous decrease in electrolyte (anolyte) conductivity charge could be due to unreacted chlorine or powered anodic partial oxidation of ethylene during operation, resulting in lowered per- hypochlorite species in the electrolyte; this to ethylene oxide (fig. S1) (full details of the TEA formance (fig. S6B). This system enables the was confirmed by using iodometric titration are available in the supplementary materials). generation of ethylene oxide in a single elec- (fig. S7 and table S1). Pt and iridium oxide/ For the TEA, we set a base electricity cost of trolyzer under ambient temperatures and pres- titantium (IrO2/Ti) anodes resist corrosion 10 ¢/kWh, which is at least twice the average sures: Ethylene, water, and electricity are the even at high chlorine concentrations during present-day industrial electricity cost (fig. S8) consumables (Eq. 7). the chlor-alkali process (31). A continuous flow (6). Sensitivity analysis reveals that the greatest

Leow et al., Science 368, 1228–1233 (2020) 12 June 2020 3of6 RESEARCH | REPORT Downloaded from http://science.sciencemag.org/

Fig. 3. Optimization of energy efficiency to reduce energy cost and and the corresponding plant-gate levelized cost as a function of current maximize technoeconomic benefit. (A) TEA showing plant-gate levelized density. Our half-cell energy efficiencies are based on our reported cost as a function of energy efficiency and renewable energy cost. potentials versus Ag/AgCl, which are not IR corrected. Additionally, we (B) Half-cell energy efficiency and the corresponding plant-gate levelized assume no losses at the cathode side, where hydrogen evolution occurs. cost as a function of Cl– concentration. (C) EDX images showing the distri- The error bars correspond to the standard deviation of three independent bution of Ir, Ti, and O on the IrO2/Ti mesh. (D) Half-cell energy efficiency measurements. on June 11, 2020

dependency of the plant-gate levelized cost is cyofthereactionbyvaryingtheelectrolyte versus Ag/AgCl and are not solution resist- on electrochemical parameters such as current concentration while operating at 300 mA/cm2. ance corrected. density and Faradaic efficiency (fig. S8; the We began at a lower Cl– concentration (0.5 M); Even at the optimal Cl– concentration, the range of values considered for each param- however, O2 evolution from water then dom- renewable electricity–based plant-gate level- eteris are provided in table S2). On the basis of inates the anodic reaction, resulting in a low ized cost remains higher than the current the current market price per ton of ethylene Faradaic efficiency of 30(±1)% and energy market price per ton of ethylene oxide and the oxide and the corresponding quantity of hydro- efficiency of 11(±1)% (Fig. 3B). As the Cl– con- corresponding quantityofhydrogen(Fig.3B). gen produced at the cathode, we determined centration increases (1.0 and 2.0 M), the po- We turned to the working electrode (catalyst) that for a current density of 300 mA/cm2,the tential decreases [5.8(±0.2) V and 4.0(±0.1) V] as another degree of freedom to decrease the – minimum energy efficiency required for the because of improved Cl oxidation kinetics potential. We prepared IrO2 deposited on Ti renewable energy–powered process to be profi- and increased electrolyte conductivity, lead- mesh (Fig. 3C and fig. S10) using a dip coat- table is ~30%. We also calculated the minimum ing to increased Faradaic efficiencies [70(±1)% ing and thermal decomposition procedure energy efficiencies required to be profitable and 67(±1)%] and half-cell energy efficiencies (33). X-ray photoelectron spectroscopy (XPS) for different electricity costs up to 20 ¢/kWh, [18(±1)% and 27(±1)%]. At 3.5 M, however, the results confirmed the presence of Ir in a +4 showing profitable regions as a function of energy efficiency was unimproved at 26(±1.9)% oxidation state (fig. S10, A to C). Scanning energy efficiency and electricity cost (Fig. because the reduced potential [3.6(±0.2) V] electron microscopy (SEM) images show the 3A). Similarly, this was also calculated for the was negated by a slight decrease in Faradaic microscale mesh structure of the IrO2-coated electrosynthesis of propylene oxide from pro- efficiency to 55(±1)%, likely because the in- Ti mesh (fig. S10D). Energy-dispersive x-ray pylene (fig. S9). creased Cl– concentration is unfavorable for spectroscopy (EDX) confirmed the presence The sensitivity analysis in fig. S8 revealed the disproportionation of Cl2 into HOCl and of Ir and O on the Ti mesh, indicating the that the plant-gate levelized cost is sensitive to HCl (Eq. 2). Thus, on the basis of the corre- loading of IrO2 on Ti (Fig. 3C). X-ray diffrac- electrochemical parameters such as Faradaic sponding plant-gate levelized costs, we de- tion (XRD) was also performed on the IrO2 efficiency and cell potential. To reduce energy termined the optimal Cl– concentration to be coatingaswellasthebareTimesh(fig.S10E). cost, we sought to increase the energy efficien- 2.0 M. In this work, all potentials are reported Additionally, transmission electron microscopy

Leow et al., Science 368, 1228–1233 (2020) 12 June 2020 4of6 RESEARCH | REPORT Downloaded from http://science.sciencemag.org/

Fig. 4. Evaluation of ethylene-to–ethylene oxide performance. (A) Half- of the CO2-to–ethylene oxide (EO) process in which the ethylene-to-EO cell potential and Faradaic efficiency of ethylene oxide over 100 hours at cell was directly supplied with the gas output from a CO2-to-ethylene MEA. 300 mA/cm2.(B) Comparison of current density, product generation rate, (D) Faradaic efficiencies of ethylene (in MEA) and ethylene oxide (in flow cell) reported operation time, Faradaic efficiency, and product specificity against as a function of the gas flow rate. For all cases, the MEA was run at state-of-the-art anodic upgrading reactions. Specificity refers to the percentage 240 mA/cm2, and the ethylene oxidation flow cell was operated at of all reacted substrate going toward the desired product. (C) Schematic 300 mA/cm2, for a duration of 1 hour.

on June 11, 2020

(TEM) images of the IrO2 were acquired (fig. On the basis of this analysis, we investigated ene oxide from CO2 (rather than ethylene) as S10, F and G). Using this catalyst, we lowered the stability of the catalyst system at the most the starting feedstock. This provides a route to the required applied potential from 3.4(±0.1) profitable current density of 300 mA/cm2, directly use renewable electricity for recycling V to 3.0(±0.1) V, thus further raising the half-cell during which portions of the electrolyte were CO2 into a valuable commodity chemical. In this 2 energy efficiency to 30(±1)% at 300 mA/cm . periodically removed for analysis and replaced integrated system, CO2 reduction to ethylene Having optimized the electrochemical sys- with fresh electrolyte. The system maintained is first performed by using a membrane electrode tem, we measured the energy efficiencies and a stable applied potential of 2.86(±0.02) V and assembly (MEA) in a gas diffusion configura- plant-gate levelized costs under different current Faradaic efficiency averaging 71(±0.6)% for tion, with O2 evolution from water as the cor- densities to determine the most economical 100 hours continuously (Fig. 4A). Post-reaction responding anodic reaction (Fig. 4C). The MEA conditions for industrial manufacturing (Fig. analysis of the anode through SEM and EDX comprises a copper nanoparticle/copper/ 3D). Faradaic efficiencies were maintained even revealed no obvious structural changes of polytetrafluoroethylene (Cu NPs/Cu/PTFE) 2 at a current density of 1 A/cm [60(±4)%]. the Ti mesh surface nor loss of IrO2 (fig. S11). cathode and an IrO2/Ti mesh anode separated However, a much higher potential of 6.5(±0.5) The method substantially outperforms other by an AEM, through which 0.1 M potassium V was required to drive the larger current, lead- reported anodic upgrading reactions in cur- bicarbonate (KHCO3) anolyte was continu- ing to a low half-cell energy efficiency [12(±1)%]. rent density, product generation rate, and re- ouslycirculated.Theoperatingcurrentdensity On the other hand, the half-cell energy effi- ported operation time while maintaining high was kept at 240 mA/cm2, and the ethylene ciency is high at 38.3(±0.1)% under 50 mA/cm2; Faradaic efficiency and ethylene oxide spec- Faradic efficiency was generally maintained thus, the electricity cost per ton of ethylene ificity (Fig. 4B). The specificity in this case is at 43 to 52% (Fig. 4D). We measured the flow oxide is at the lowest. However, the high cap- 95%; we did not observe the conversion of rate of the output gas using a flow meter at ital cost associated with electrolyzer surface ethylene to other products (such as CO2). This the cathode gas outlet, and the output gas area resulted in an uneconomical plant-gate high specificity is important in an industrial was directly sparged into the anolyte of the levelized cost. The plant-gate levelized cost is process because the ethylene will likely be ethylene-to–ethylene oxide flow cell (operated the lowest at 300 mA/cm2,withgoodenergy continuously recirculated to maximize usage. at 300 mA/cm2) without further purification. efficiency of 30(±1)% and acceptably low cap- Last, we sought to develop an integrated With this method, we achieved a Faradaic ital costs. system to perform the electrosynthesis of ethyl- efficiency of 45% toward ethylene oxide under

Leow et al., Science 368, 1228–1233 (2020) 12 June 2020 5of6 RESEARCH | REPORT

a gas flow rate of 6 standard cubic centimeters 4. J. Zheng, S. Suh, Nat. Clim. Chang. 9, 374–378 (2019). 32. “Market Analytics: Propylene Oxide - 2018,” Markets & per minute (sccm) (Fig. 4D), despite the pres- 5. P. De Luna et al., Science 364, eaav3506 (2019). Profitability (Nexant, 2018). 6. M. Jouny, W. Luc, F. Jiao, Ind. Eng. Chem. Res. 57, 2165–2177 – enceofothereasilyoxidizablegasessuchas 33. W. Luc, J. Rosen, F. Jiao, Catal. Today 288,79 84 (2017). (2018). 34. Y. Huang, X. Chong, C. Liu, Y. Liang, B. Zhang, Angew. Chem. H2 and CO relative to ethylene (23% H2,12% 7. C. Xia, Y. Xia, P. Zhu, L. Fan, H. Wang, Science 366, 226–231 Int. Ed. 57, 13163–13166 (2018). CO, and 12% ethylene) (fig. S12 and tables S3 (2019). 35. C. Huang, Y. Huang, C. Liu, Y. Yu, B. Zhang, Angew. Chem. Int. – 8. R. F. Service, Science 365, 1236 1239 (2019). Ed. 58, 12014–12017 (2019). and S4). Oxidation of these gases requires ’ 9. S. Rebsdat, D. Mayer, in Ullmann s Encyclopedia of Industrial 36. Y. Lum et al., Nat. Catal. 3,14–22 (2020). direct contact with the anode, whereas ethyl- Chemistry. (Wiley, 2001). ene oxidation is mediated by the extended 10. World Program (WP), “Ethylene,” WP Report ACKNOWLEDGMENTS (SRI Consulting, 2009). heterogeneous:homogeneous interface and We thank D. Kopilovic and R. Wolowiec for their kind technical 11. World Petrochemicals Program (WP), “Ethylene oxide,” WP assistance. Funding: This material is based on work supported thus occurs in the bulk electrolyte at a much Report (SRI Consulting, 2009). by the Ontario Ministry of Colleges and Universities (grant ORF- 12. A. Boulamanti, J. A. Moya, Energy Efficiency and GHG higher rate. The Faradaic efficiency toward RE08-034), Natural Sciences and Engineering Research Council Emissions: Prospective Scenarios for the Chemical and ethylene oxide was reduced at a higher gas (NSERC) of Canada (grant RGPIN-2017-06477), Canadian Institute Industry (Publications Office of the European for Advanced Research (CIFAR) (grant FS20-154 APPT.2378), flow rate owing to lowered ethylene concen- Union, 2017). and University of Toronto Connaught Fund (grant GC 2012-13). trationintheMEAoutputstream(fig.S12and 13. S. T. Wismann et al., Science 364, 756–759 (2019). D.S. acknowledges the NSERC E. W. R. Steacie Memorial tables S3 and S4). However, decreasing the 14. K. M. Van Geem, V. V. Galvita, G. B. Marin, Science 364, Fellowship. Author contributions: E.H.S. supervised the project. – 734 735 (2019). W.R.L, Y.L., and E.H.S. conceived the idea and designed the flow rate even further (3 sccm) resulted in a – 15. H. G. Cha, K.-S. Choi, Nat. Chem. 7, 328 333 (2015). experiments. W.R.L and Y.L. carried out all the experimental work. lowered Faradaic efficiency toward ethylene 16. N. Jiang, B. You, R. Boonstra, I. M. Terrero Rodriguez, Y. Sun, A.O. fabricated the IrO2-coated Ti mesh electrodes and assisted in ACS Energy Lett. 1, 386 –390 (2016). in the MEA. This reduces the ethylene supply the CO2-to–ethylene oxide conversion experiments. Y.W. available for conversion in the flow cell, re- 17. B. You, N. Jiang, X. Liu, Y. Sun, Angew. Chem. Int. Ed. 55, performed SEM measurements. D.-H.N. performed the XRD 9913–9917 (2016). measurements. B.C. carried out the TEM measurements. J.W. sulting in a drop in the Faradaic efficiency 18. T. Li, Y. Cao, J. He, C. P. Berlinguette, ACS Cent. Sci. 3,

carried out the XPS measurements. T.-T.Z., F.L., and D.S. Downloaded from toward ethylene oxide. Thus, both concentra- 778–783 (2017). contributed to data analysis and manuscript editing. W.R.L, Y.L., tion and molar quantity of ethylene in the 19. R. S. Sherbo, R. S. Delima, V. A. Chiykowski, B. P. MacLeod, and E.H.S. cowrote the manuscript. All authors discussed the – MEA output stream are important determi- C. P. Berlinguette, Nat. Catal. 1, 501 507 (2018). results and assisted during the manuscript preparation. 20. J. Zheng et al., Adv. Funct. Mater. 27, 1704169 (2017). Competing interests: W.R.L, Y.L., and E.H.S. have filed provisional nants for the Faradaic efficiency toward ethyl- 21. D. Liu et al., Nat. Commun. 10, 1779 (2019). patent application no. 63/002,653 regarding the electrosynthesis ene oxide in the flow cell. 22. Y. Kwon, Y. Birdja, I. Spanos, P. Rodriguez, M. T. M. Koper, ACS of oxiranes. Data and materials availability: All experimental data This demonstration shows the viability of Catal. 2, 759–764 (2012). are available in the main text or the supplementary materials. 23. C. Dai et al., J. Catal. 356,14–21 (2017). – http://science.sciencemag.org/ an integrated system for complete CO2-to 24. A. Winiwarter et al., Energy Environ. Sci. 12, 1055–1067 (2019). ethylene oxide conversion. Further improve- 25. M. Rafiee, K. C. Miles, S. S. Stahl, J. Am. Chem. Soc. 137, SUPPLEMENTARY MATERIALS – ments are expected by optimizing the ethylene 14751 14757 (2015). science.sciencemag.org/content/368/6496/1228/suppl/DC1 Faradaic efficiency and single-pass conversion 26. E. J. Horn et al., Nature 533,77–81 (2016). Materials and Methods 27. M. Rafiee, F. Wang, D. P. Hruszkewycz, S. S. Stahl, J. Am. Supplementary Text in the MEA. Chem. Soc. 140,22–25 (2018). Figs. S1 to S12 28. C. L. McCABE, J. C. Warner, J. Am. Chem. Soc. 70, 4031–4034 Tables S1 to S4 REFERENCES AND NOTES (1948). References (37–39) – 1. U.S. Energy Information Administration (EIA), “Energy use in 29. M. Eigen, K. Kustin, J. Am. Chem. Soc. 84, 1355 1361 (1962). industry,” in Use of Energy Explained (EIA, 2019). 30. W. Tong et al., Nat. Energy (2020). 14 October 2019; resubmitted 26 January 2020 2. S. Chu, Y. Cui, N. Liu, Nat. Mater. 16,16–22 (2016). 31. R. K. B. Karlsson, A. Cornell, Chem. Rev. 116, 2982–3028 Accepted 1 April 2020 3. Z. W. Seh et al., Science 355, eaad4998 (2017). (2016). 10.1126/science.aaz8459

on June 11, 2020

Leow et al., Science 368, 1228–1233 (2020) 12 June 2020 6of6 Chloride-mediated selective electrosynthesis of ethylene and propylene oxides at high current density Wan Ru Leow, Yanwei Lum, Adnan Ozden, Yuhang Wang, Dae-Hyun Nam, Bin Chen, Joshua Wicks, Tao-Tao Zhuang, Fengwang Li, David Sinton and Edward H. Sargent

Science 368 (6496), 1228-1233. DOI: 10.1126/science.aaz8459

Charging into epoxides Ethylene oxide is a strained, reactive molecule produced on a vast scale as a plastics precursor. The current

method of synthesis involves the direct reaction of ethylene and oxygen at high temperature, but the original protocol Downloaded from relied on the reduction of chlorine to produce a chlorohydrin intermediate. Leow et al. report a room temperature method that returns to the chlorine route but uses electrochemistry to generate it catalytically from chloride (see the Perspective by Barton). This efficient process uses water in place of oxygen and can be integrated with the electrochemical generation of ethylene from carbon dioxide. Propylene oxide can be produced using the same method. Science, this issue p. 1228; see also p. 1181 http://science.sciencemag.org/

ARTICLE TOOLS http://science.sciencemag.org/content/368/6496/1228

SUPPLEMENTARY http://science.sciencemag.org/content/suppl/2020/06/10/368.6496.1228.DC1 MATERIALS

on June 11, 2020 RELATED http://science.sciencemag.org/content/sci/368/6496/1181.full CONTENT

REFERENCES This article cites 31 articles, 6 of which you can access for free http://science.sciencemag.org/content/368/6496/1228#BIBL

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Use of this article is subject to the Terms of Service

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. The title Science is a registered trademark of AAAS. Copyright © 2020 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S. Government Works